• Current research
    Projects (2)
    Archived project
    Statistical Power to Detect Individual Differences in Change. Latent Growth Curve Models. Effect Sizes. Reliability.
    Research Items
    The main objective of "Lifebrain" is to identify the determinants of brain, cognitive and mental (BCM) health at different stages of life. By integrating, harmonising and enriching major European neuroimaging studies across the life span, we will merge fine-grained BCM health measures of more than 5,000 individuals. Longitudinal brain imaging, genetic and health data are available for a major part, as well as cognitive and mental health measures for the broader cohorts, exceeding 27,000 examinations in total. By linking these data to other databases and biobanks, including birth registries, national and regional archives, and by enriching them with a new online data collection and novel measures, we will address the risk factors and protective factors of BCM health. We will identify pathways through which risk and protective factors work and their moderators. Exploiting existing European infrastructures and initiatives, we hope to make major conceptual, methodological and analytical contributions towards large integrative cohorts and their efficient exploitation. We will thus provide novel information on BCM health maintenance, as well as the onset and course of BCM disorders. This will lay a foundation for earlier diagnosis of brain disorders, aberrant development and decline of BCM health, and translate into future preventive and therapeutic strategies. Aiming to improve clinical practice and public health we will work with stakeholders and health authorities, and thus provide the evidence base for prevention and intervention.
    Well-being is often relatively stable across adulthood and old age, but typically exhibits pronounced deteriorations and vast individual differences in the terminal phase of life. However, the factors contributing to these differences are not well understood. Using up to 25-year annual longitudinal data obtained from 4,404 now-deceased participants of the nationwide German Socio-Economic Panel Study (SOEP; age at death: M = 73.2 years; SD = 14.3 years; 52% women), we explored the role of multi-indicator constellations of sociodemographic variables, physical health and burden factors, and psychosocial characteristics. Expanding earlier reports, structural equation model (SEM) trees allowed us to identify profiles of variables that were associated with differences in the shape of late-life well-being trajectories. Physical health factors were found to play a major role for well-being decline, but in interaction with psychosocial characteristics such as social participation. To illustrate, for people with low social participation, disability emerged as the strongest correlate of differences in late-life well-being trajectories. However, for people with high social participation, whether or not an individual had spent considerable time in the hospital differentiated high versus low and stable versus declining late-life well-being. We corroborated these results with variable importance measures derived from a set of resampled SEM trees (so-called SEM forests) that provide robust and comparative indicators of the total interactive effects of variables for differential late-life well-being. We discuss benefits and limitations of our approach and consider our findings in the context of other reports about protective factors against terminal decline in well-being.
    Question - What does a replication of a null result mean to you?
    Audrey, let's assume the supervisors of both researchers A and yourself are ignorant of this issue and will only prolong your contract if you find evidence to reject X.
    Question - What is the best practice to deal with limited sample size?
    For the special case of longitudinal designs, Timo von Oertzen and I presented a theory of power equivalent designs. We explain how different design parameters can be traded against each other to hold a certain desired power constant. For example, this allows trading the number of measurement time points against the reliability of the instrument or can be used to trade sample size for number of time points. Virtually, any parameter can be traded against any other. A more general theory for all types of linear Gaussian models was mathematically laid out by Timo von Oertzen before.
    von Oertzen, T., & Brandmaier, A. M. (2013). Optimal Study Design With Identical Power: An Application of Power Equivalence to Latent Growth Curve Models. Psychology and Aging, Vol 28(2), 414-428.
    Oertzen, T. (2010). Power equivalence in structural equation modelling. British Journal of Mathematical and Statistical Psychology, 63(2), 257-272.
    Question - What does a replication of a null result mean to you?
    In fact, I had a typo before. You are supposed to not believe in the null hypothesis and try to replicate the study to reject it.
    Question - What is your favorite Structural Equation Modeling program?
    Rob, I agree. It's a good idea to compare results across programs. It's not only double-checking programs but also double-checking your own programming. Btw., Onyx can export to M+, lavaan, and OpenMx, which is handy for crosschecking results.
    The brain-derived neurotrophic factor (BDNF) promotes activity-dependent synaptic plasticity, and contributes to learning and memory. We investigated whether a common Val66Met missense polymorphism (rs6265) of the BDNF gene is associated with individual differences in cognitive decline (marked by perceptual speed) in old age. A total of 376 participants of the Berlin Aging Study, with a mean age of 83.9 years at first occasion, were assessed longitudinally up to 11 times across more than 13 years on the Digit-Letter task. Met carriers (n = 123, 34%) showed steeper linear decline than Val homozygotes (n = 239, 66%); the corresponding contrast explained 2.20% of the variance in change in the entire sample, and 3.41% after excluding individuals at risk for dementia. These effects were not moderated by sex or socioeconomic status. Results are consistent with the hypothesis that normal aging magnifies the effects of common genetic variation on cognitive functioning. (PsycINFO Database Record (c) 2014 APA, all rights reserved).
    Question - How many times must I repeat one experiment to find the appropriate probability of occurrence?
    I realize that many questions in researchgate ask for the "best" method without a definition of what the authors of the question mean with "best". For a given statistical definition of "best", an analytic answer can typically be calculated, at least for coin flips. Eik has already pointed in a direction that is commonly accepted.
    Question - How can I perform time series data similarity measures and get a significance level (p-value)?
    To answer this question, you need to formalize dissimilarity. A simple dissimilarity measure could be a geometrical distance, e.g., the Euclidean distance between two time-series. This dissimilarity has certain properties that can be advantageous or disadvantageous depending on your research question. For example, if your dissimilarity is supposed to be invariant to scaling, Euclidean distance is typically a poor choice (unless you normalize but this imposes additional assumptions). Among other approaches, dissimilarity can be formalized in terms of shape, based on parameters of a hypothesized generating model, or complexity of the time series. The R package PDC provides complexity-based dissimilarity calculation and clustering, and also provides p values for a null hypothesis of identical underlying generating permutation distributions. The R package TSclust was recently updated and provides (among PDC) a number of approaches to time series dissimilarities.
    Andreas M.…
    Question - What are the best longitudinal statistical analyses to be applied to a small sample of patients with dementia?
    For future study design planning, you might be interested in the comparison of power-equivalent research designs. My colleague Timo von Oertzen and I have recently published a paper on how different design factors (e.g., number of measurement occasions, total study time span, indicator reliability) can be traded against each other while maintaining the same power in longitudinal studies.
    Question - Who is the Father of Information Technology?
    Claude Shannon for his mathematical theory of communication.
    Question - What is the best way to analyse location data?
    There are many ways to formalize difference between individuals. In a recent study, we used roaming entropy to quantify differences in the roaming behavior of mice and found that these differences are correlated with neurogenesis in the hippocampus. Essentially, roaming entropy is the entropy of the spatial distribution representing the whereabouts of a mouse. More information can be found here:
    Freund, J., Brandmaier, A. M., Lewejohann, L., Kirste, I., Kritzler, M., Krüger, A., Sachser, N., Lindenberger, U., & Kempermann, G. (2013). Emergence of individuality in genetically identical mice. Science, 340(6133), 756-759. doi:10.1126/science.1235294
    Top co-authors